Abstract: LiDAR-Camera fusion-based 3D object detection is one of the main visual perception tasks in autonomous driving, facing the challenges of small targets and occlusions. Image semantics are ...